Intel shared the first details about its upcoming Meteor Lake platform, which will integrate Arc graphics and a VPU to handle AI inference workloads.
Intel has shared the first details of the upcoming Meteor Lake platform, which will debut later this year with a big focus on client-side AI performance. AI is the hot topic in computing right now, and Intel is betting hard to make sure it can deliver capable AI experience on client devices so you don’t have to rely on the cloud for all the workloads and use cases that are starting to show up.
Meteor Lake uses a chiplet design
Source: Intel
Meteor Lake is the next generation of Intel CPUs, and it’s the first one to be based on the Intel 4 process while also leveraging Intel’s Foveros packaging technology in a chiplet-style design, where different components of the processor are built in separate tiles (or chiplets). This chiplet design allows for advanced power management, and it also means each of the tiles can be individually targeted for specific workloads.
One of these chiplets is, of course, the CPU, which has been developed with a major focus on power efficiency, retaining the hybrid architecture we’ve seen in 12th-gen and 13th-gen processors. There’s also a brand-new GPU, with Intel finally bringing Arc graphics to an integrated architecture, including support for things like ray tracing and XeSS in a low-power design.
The big focus for Intel, however, is the new AI engine, or the VPU (vision processing unit), built into the processor. This is what Intel hopes will push client-side AI workloads forward, significantly reducing the compute requirements of AI inferencing.
Intel is going all-in on AI
By now, everyone has probably heard of AI, which is showing up in many forms, whether that’s a conversational AI like ChatGPT based on a large language model, but also image generation with tools like Midjourney and Stable Diffusion. These workloads are typically handled in the cloud, however, which can mean increased costs for software vendors, especially as these tools become more advanced and have higher compute requirements, as well as privacy concerns for users. And with features like Windows Copilot, AI is going to be everywhere, so Intel doesn’t want to miss out.
That’s where the VPU in the Meteor Lake platform comes in. This specialized processor is designed to handle specific AI workloads much more efficiently than a general-purpose CPU or even a GPU could, thus enabling this kind of workload to run locally on a computer. The VPU is designed for sustained AI workloads, such as applying effects like background blur, while the GPU can be useful for AI infused into render pipelines, and the CPU can be used for lightweight single-inference tasks. All these components can also be leveraged at the same time for workloads that can make use of the strengths of each one.
One example Intel provided was running a Stable Diffusion image generation in GIMP on the Meteor Lake platform, where the computer generated a complex image based on a text description, leveraging the CPU, GPU, and VPU, taking about 20 seconds to do so. Then, it’s also possible to run a Super Resolution tool to increase the resolution of the image, and that runs exclusively on the VPU, providing a higher-resolution version of the image in just a few seconds. The enhanced efficiency and performance of the VPU also means effects like background blur can work much better, with improved edge detection, while still using less power.
For developers, Intel is also leveraging platforms and tools like ONNX, OpenVINO, and Windows DirectML to bring these capabilities to existing apps, including GIMP, Blender, Audacity, OBS, and Adobe apps. Intel is playing to its strengths, particularly the robust ecosystem of thousands of x86 apps and platforms that are already built around its processors. This is a major advantage over competitors like Qualcomm, which struggles with app compatibility on Windows.
Intel has yet to share details about the availability of Meteor Lake processors, or specific specs for the CPU and GPU, but it’s clear that the focus is on AI, not just for Intel, but the whole industry.